intent-driven-development
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill includes explicit security guardrails, such as Operating Rule 8, which forbids the inclusion of real secrets, credentials, tokens, or private keys in any generated criteria or artifacts, requiring the use of redacted or synthetic values.
- [SAFE]: Operating Rule 9 prevents the execution of destructive tests, migrations, or operations against live production data without explicit user authorization and a safe environment.
- [SAFE]: The instructions emphasize least-privilege principles by instructing the agent not to modify repository files, branches, or commits unless specifically requested by the user (Operating Rule 6).
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for potentially untrusted data by reading repository files and user-provided specifications.
- Ingestion points: Local repository files, technical documentation, and user-provided PRDs.
- Boundary markers: The workflow instructions provide clear logical separation between 'discovered facts' (code-based) and 'product constraints' (user-based) to prevent misinterpretation of requirements.
- Capability inventory: Contextual file reading and potential file writing (if requested).
- Sanitization: Mandatory redaction of sensitive payloads and validation of observable outcomes are required before outputting results.
Audit Metadata